Deep Learning-Based Direction-of-Arrival Estimation in Automotive MIMO Radars under Multipath Propagation

Introduction. Today, millimeter-wave radars are a key component of driver-assistance systems. However, in urban scenarios, the performance of conventional methods for estimating the direction of arrival of signals (DOA) is degraded by multipath effects, which can lead to ghost targets. This article describes an algorithm for classifying signals caused by single reflection and multipath propagation and estimating their DOA.
Aim. To develop and investigate experimentally a method for estimating the angular coordinates of objects in MIMO radars in the presence of multipath propagation.
Materials and methods. A two-stage signal processing algorithm was proposed. In the first stage, a convolutional neural network classifies the signal scenario as one of the following types: direct path, multipath, or multiple target. In the second stage, the direction of arrival is estimated by the standard MUSIC method and a modified MUSIC method based on reconstruction of the signal correlation matrix. Training and validation were performed on a hybrid dataset consisting of data generated using a signal propagation model and actual measurements obtained by a TI AWR1843 MIMO radar.
Results. The proposed algorithm can be used to classify signal propagation scenarios with high accuracy and to estimate the angular coordinates of objects. The proposed method was tested on an open dataset. The results show that the classification results are consistent with the mathematical model of signal propagation.
Conclusion. The proposed approach is effective for multipath detection. The proposed algorithm allows the robustness of MIMO radar scene estimation to be improved.

1. Patole S. M., Torlak M., Wang D., Ali M. Automotive Radars: a Review of Signal Processing Techniques. IEEE Signal Processing Magazine. 2017, vol. 34, iss. 2, pp. 22–35. doi: 10.1109/MSP.2016.2628914

2. Robey F. C., Coutts S., Weikle D., McHarg J. C., Cuomo K. MIMO Radar Theory and Experimental Results. Conf. Record of the Thirty-Eighth Asilomar Conference on Signals, Systems and Computers, Pacific Grove, USA, 07–10 Nov. 2004. IEEE, 2004, pp. 300–304. doi: 10.1109/ACSSC.2004.1399141

3. Biallawons O., Ender J. H. G. Multipath Detection by using Space-Space Adaptive Processing (SSAP) with MIMO Radar. Intern. Conf. on Radar, Brisbane, Australia, 27–31 Aug. 2018. IEEE, 2018, pp. 1–4. doi: 10.1109/RADAR.2018.8557221

4. Longman O., Villeval Sh., Bilik I. Multipath Ghost Targets Mitigation in Automotive Environments. IEEE Radar Conf. (RadarConf21), Atlanta, USA, 07–14 May 2021. IEEE, 2021, pp. 1–5. doi: 10.1109/RadarConf2147009.2021.9455253

5. Yang R., Hu Y., Sun Sh., Zhang Y. D. Advancing Subspace Representation Learning for DOA Estimation using Sparse Arrays. Proc. of the 59th Asilomar Conf. on Signals, Systems, and Computers, Pacific Grove, California, USA, 26–29 Oct. 2025. IEEE, 2025, pp. 1–15. doi: 10.1109/IEEECONF67917.2025.11443901

6. Lee Ch. Y., Hasegawa H., Gao Sh. Complex-Valued Neural Networks: A Comprehensive Survey. IEEE/CAA J. of Automatica Sinica. 2022, vol. 9, iss. 8, pp. 1406–1426. doi: 10.1109/JAS.2022.105743

7. Skolnik M. I. Radar Handbook. 2nd ed. New York, McGraw-Hill, 1990, 1328 p.

8. Waldschmidt C., Hasch J., Menzel W. Automotive Radar – From First Efforts to Future Systems. IEEE J. of Microwaves. 2021, vol. 1, iss. 1, pp. 135–148. doi:10.1109/JMW.2020.3033616

9. Simonyan K., Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition. 3rd Intern. Conf. on Learning Representations (ICLR 2015), San Diego, California, USA, 7–9 May 2015. IEEE, 2015, pp. 1–14.

10. Spielman D., Paulraj A., Kailath T. Performance Analysis of the MUSIC Algorithm. IEEE Intern. Conf. on Acoustics, Speech, and Signal Processing, Tokyo, Japan, 07–11 April 1986. IEEE, 1986, pp. 1909–1912. doi: 10.1109/ICASSP.1986.1168872

11. Changgan Sh., Yumin L. An Improved forward/Backward Spatial Smoothing Root-MUSIC Algorithm Based on Signal Decorrelation. IEEE Workshop on Advanced Research and Technology in Industry Applications, Ottawa, ON, 29–30 Sept. 2014. IEEE, 2014, pp. 1252–1255. doi: 10.1109/WARTIA.2014.6976509

12. Wang Y., Si W., Wang K. A Real-Valued MUSIC Algorithm with Forward/Backward Technique. Intern. Applied Computational Electromagnetics Society Symp., Firenze, Italy, 26–30 March 2017. IEEE, 2017, pp. 1–3. doi: 10.23919/ROPACES.2017.7916026

13. Alam A. M., Ayna C. O., Biswas S., Rogers J. T., Ball J. E., Gurbuz A. C. Deep Learning-Based Covariance Matrix Reconstruction for DOA Estimation. IEEE Radar Conf., Denver, CO, USA, 06–10 May 2024. IEEE, 2024, pp. 1–6. doi: 10.1109/RadarConf2458775.2024.10548988

14. Zhao Y., Liu J., Fan X., Cao H. Direction-of-Arrival Estimation Using Deep Learning With Covariance Matrix Reconstruction Under Limited Snapshots. IET Electronic Let. 2025, vol. 61, iss. 1, art. no. e70373. doi: 10.1049/ell2.70373

15. Gao X., Xing G., Roy S., Liu H. RAMP-CNN: A Novel Neural Network for Enhanced Automotive Radar Object Recognition. IEEE Sensor J. 2021, vol. 21, iss. 4, pp. 5119–5132. doi: 10.1109/JSEN.2020.3036047

16. AWR1843BOOST and IWR1843BOOST Single-Chip mmWave Sensing Solution User’s Guide (Rev. B). Available at: https://www.ti.com/tool/AWR1843BOOST#tech-docs (accessed: 10.02.2026)